Constructing a statistical shape model from two-dimensional or three-dimensional data

Abstract

A statistical shape model is built by automatically establishing correspondence between a set of two dimensional shapes or three dimensional shapes. A parameterization of each shape is determined and, a statistical shape model is built using the parameterization. An objective function is used to provide an output which indicates the quality of the statistical shape model. By performing these steps repeatedly for different parameterizations and comparing the quality of the resulting statistical shape models using output of the objective function to determine which parameterization provides the statistical shape model having the best quality, the output of the objective function is a measure of the quantity of information required to code the set of shapes using the statistical shape model.

Description

BACKGROUND[0001]1. Technical Field[0002]The present invention relates to a statistical shape model, and to the parameterization of a set of shapes used for the statistical shape model.[0003]2. Related Art[0004]Statistical models of shape have been used for some time to provide automated interpretation of images. See, for example, Cootes, T, A. Hill, and C. Taylor, The use of Active shape models for locating structures in medical images. Image and Vision Computing, 1994, 12: p. 355-366. The basic idea used by the models is to establish, from a training set, a pattern of “legal” variation in the shapes and spatial relationships of structures on a given class of images (the class of images may be for example face images, or hand images, etc.). Statistical analysis is used to give an efficient parametensation of the pattern of legal variation, providing a compact representation of shape. The statistical analysis also provides shape constraints which are used to determine whether the sha...

AI Technical Summary

Benefits of technology

[0018]It is an object of an exemplary embodiment of the present invention to provide a method of parameterization

Problems solved by technology

One of the main drawbacks to statistical shape models is the need, during training, to establish dense correspondence between shape boundaries for a reasonably large set of example images.

If ‘correct’ correspondences are not established, an inefficient model of shape can result, leading to difficulty in defining shape constraints.

In other words, the model will not correctly determine whether the shape of a hypothesised structure in an analysed image represents a plausible example of the object class of interest.

However, there are several disadvantages associated with manually defining landmarks.

Firstly, in a general a large number of images must be annotated in order to generate an accurate model, and manually defining landmarks for each image is very time consuming.

A second disadvantage is that manually defining the landmarks unavoidably involves an element of subjective judgement when determining exactly where to locate each landmark, and this will lead to some distortion of the model.

The disadvantages are exacerbated when manually defining landmarks for 3-D images, since the number of landmarks per image increases significantly.

An inappropriate choice

Images